WASHINGTON - Hoping to set the record straight on fair-lending, the American Bankers Association has released an exhaustive analysis detailing problems with statistical studies.
The ABA recruited leading economists from the banking agencies, academia, and industry to write its first broadside in the battle of lending discrimination studies.
"The ABA has published this compendium to educate observers of the banking industry regarding the importance of doing this sort of research correctly, and to highlight the difficulties facing researchers in this field," ABA economist Mike ter Maat said.
Regulators, economists, and community activists have released more than a dozen comprehensive lending bias studies during the past year. Banking industry advocates have criticized many of the reports, saying they contain statistical flaws.
Federal Reserve Board Governor Lawrence B. Lindsey wrote in the introduction of the 175-page Fair Lending Analysis that statistical models can't provide conclusive information on individual borrowers. Instead, they provide "educated bets" on whether discrimination occurred.
"Surely this approach should not form the basis of identifying victims of illegal discrimination," he said.
The ABA book is divided into four parts - redlining, loan approvals, loan defaults, and regulatory/judicial actions.
The redlining section concentrates on the use of Home Mortgage Disclosure Act data to identify discrimination against predominantly low- income and minority neighborhoods.
George Washington University Prof. Anthony Yezer, who edited the book, wrote that the five studies in this section prove that HMDA data, even when augmented with census tract information, produce misleading results. The problem: HMDA doesn't measure loan demand.
The ABA also criticized loan approval studies, saying they can't identify marketwide problems because each bank has different underwriting standards. Also it said the studies only work for individual banks when the researchers have complete access to the bank's records.
The book attacks default rate studies, saying these reviews are misleading unless researchers know the credit history of each borrower.
Finally, the ABA provides a series of conflicting essays on whether statistics have any role in fair-lending enforcement. It concluded that the courts will decide the validity of these studies in enforcement matters. Essays by three banking agency economists recognize that statistical studies are effective only at identifying banks that deserve further scrutiny.
Copies of the book are available by calling 202-663-5354.